Background Long-term physiological variations, such as seasonal variations, affect the screening efficiency at medical checkups. This study examined the seasonal variation in liver function tests using recently described data-mining methods. Methods The ‘latent reference values’ of aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), gamma-glutamyltransferase ( γGT), cholinesterase (ChE) and total bilirubin (T-Bil) were extracted from a seven-year database of outpatients (aged 20–79 yr; comprising approximately 1,270,000 test results). After calculating the monthly means for each variable, the time-series data were separated into trend and seasonal components using a local regression model (Loess method). Then, a cosine function model (cosinor method) was applied to the seasonal component to determine the periodicity and fluctuation range. A two-year outpatient database (215,000 results) from another hospital was also analysed to confirm the reproducibility of these methods. Results The serum levels of test results tended to increase in the winter. The increase in AST and ALT was about 6% in men and women, and was greater than that in ChE, ALP (in men and women) and γGT (in men). In contrast, T-Bil increased by 3.6% (men) and 5.0% (women) in the summer. The total protein and albumin concentrations did not change significantly. AST and ALT showed similar seasonal variation in both institutions in the comparative analysis. Conclusions The liver function tests were observed to show seasonal variations. These seasonal variations should therefore be taken into consideration when establishing either reference intervals or cut-off values, which are especially important regarding aminotransferases.